 |
 |
Hi Emmanuel, I am not exactly an expert, so others may have simpler and/or more correct approaches, but I have encountered similar problems, and what I suppose you could do is to make a model with all your events as one condition and modulate that with event length, then go into the design matrix (SPM.xX.X)and pick up the vector with the modulated responses and add that as an extra regressor to a new design matrix where you have modelled each condition separately. Of course you need to make sure that this regressor is orthogonal to your main effects regressors which it may not be if the length variation is systematically different between conditions. In the end it all comes down to whether your conditions and your confounds are separable. I hope this was a help. Mikkel Mikkel Wallentin, Ph.d. CFIN Center for Functionally Integrative Neuroscience Aarhus University Hospital Building 30, Noerrebrogade 8000 Aarhus C, Denmark Phone: +4589494380 Fax: +4589494400 http://www.cfin.au.dk ----- Original Message ----- From: "E A Stamatakis" <[log in to unmask]> To: <[log in to unmask]> Sent: Monday, December 18, 2006 3:39 PM Subject: [SPM] Modulators-between & within conditions variance > Dear experts, > > I have run an fMRI experiment with auditory stimuli that are all under > 1s long. I have modelled these as events (7 conditions A,B,C,D, E,F & G) > in a first level model. Some conditions contain significantly longer > stimuli than others. I included the stimulus length as a modulator for > each condition but from previous messages on the list I understand > that this will covary out length differences within a condition but not > between conditions. > > So as I understand this a comparison of conditions A and B (A-B or B-A) > will not show me differences between A & B having co-varied out the > effect of length. > > There has been a suggestion on list to model all events as 1 condition > type and to then model event types as modulators. In this case all event > types will be modelled in one column and there will be 8 modulators, the > first 7 indicating whether each event is of type A, B, C,D, E,F or G and > the 8th modulator will be length. The columns indicating trial type > would be modelled with 1s and -1s. Is it the case that A-B in this type > of model will show the differences between the 2 conditions having > co-varied out the effect of length? > > Effectively I am trying to find out whether there is a significant > effect once the confound has been removed. > > Many thanks for your help, > > Emmanuel A Stamatakis PhD Tel: +44 (0) 1223 766454 > Department of Experimental Psychology Fax: +44 (0) 1223 766452 > University of Cambridge > Downing Street > Cambridge CB2 3EB > U.K. http://csl.psychol.cam.ac.uk >